This paper introduces a novel visual and interactive clustering tool. It relies on a dimensionality reduction technique to allow for a 2D representation of the data and associated clustering, set initially in an unsupervised fashion. Its main contribution is to enable iterative updates of both the 2D projection and clustering. Using appropriate controls, the user may thus inject his or her preferences, and visualize the induced change in real time. The involved dimensionality reduction technique follows a physical metaphor, that facilitates the tracking of changes by the user. The practical interest of the tool is illustrated by an example.